Update app.py

app.py

@@ -2,6 +2,7 @@ from dotenv import load_dotenv
 import os
 import gradio as gr
 from PyPDF2 import PdfReader
+import google.generativeai as genai
 from langchain.text_splitter import CharacterTextSplitter
 from langchain_google_genai import GoogleGenerativeAIEmbeddings, ChatGoogleGenerativeAI
 from langchain_community.vectorstores import FAISS
@@ -75,6 +76,43 @@ class PDFChatBot:
 
         return raw_text, processed_count
 
+    def get_pdf_text_via_gemini(self, pdf_files):
+        """使用 Gemini 2.0 Flash 直接解析 PDF 文字（透過 Files API）。"""
+        api_key = _get_api_key()
+        if not api_key:
+            return "", 0
+
+        genai.configure(api_key=api_key)
+        model = genai.GenerativeModel("gemini-2.0-flash-exp")
+
+        raw_text = ""
+        processed_count = 0
+
+        if not pdf_files:
+            return raw_text, processed_count
+
+        if not isinstance(pdf_files, list):
+            pdf_files = [pdf_files]
+
+        for pdf_file in pdf_files:
+            try:
+                pdf_path = pdf_file.name if hasattr(pdf_file, 'name') else pdf_file
+                uploaded = genai.upload_file(pdf_path)
+                prompt = (
+                    "請從此 PDF 中提取可讀文字，按頁面順序輸出純文字。"
+                )
+                resp = model.generate_content([uploaded, prompt])
+                text = resp.text or ""
+                if text.strip():
+                    raw_text += text + "\n"
+                    processed_count += 1
+                    self.processed_files.append(os.path.basename(pdf_path))
+            except Exception as e:
+                print(f"使用Gemini解析PDF時發生錯誤：{str(e)}")
+                continue
+
+        return raw_text, processed_count
+
     def get_text_chunks(self, text):
         """將文字分割成區塊進行處理"""
         text_splitter = CharacterTextSplitter(
@@ -193,7 +231,7 @@ class PDFChatBot:
         except Exception as e:
             return f"處理問題時發生錯誤：{str(e)}"
 
-    def process_pdfs(self, pdf_files, progress=gr.Progress()):
+    def process_pdfs(self, pdf_files, progress=gr.Progress(), use_gemini=False):
         """處理PDF文件"""
         if not pdf_files:
             return "請上傳至少一個PDF文件。", ""
@@ -203,7 +241,10 @@ class PDFChatBot:
 
         # 提取文字
         progress(0.2, desc="提取PDF文字內容...")
-        raw_text, processed_count = self.get_pdf_text(pdf_files)
+        if use_gemini:
+            raw_text, processed_count = self.get_pdf_text_via_gemini(pdf_files)
+        else:
+            raw_text, processed_count = self.get_pdf_text(pdf_files)
 
         if not raw_text.strip():
             return "無法從PDF文件中提取到文字。", ""
@@ -300,8 +341,8 @@ class PDFChatBot:
 bot = PDFChatBot()
 
 # Gradio 接口函數
-def upload_and_process(files, progress=gr.Progress()):
-    return bot.process_pdfs(files, progress)
+def upload_and_process(files, use_gemini=False, progress=gr.Progress()):
+    return bot.process_pdfs(files, progress, use_gemini)
 
 def ask_question(question, history, temperature, max_tokens, search_k):
     if not question.strip():
@@ -350,6 +391,20 @@ def load_existing_data():
     else:
         return "❌ 沒有找到已處理的資料。", ""
 
+def set_api_key(api_key: str):
+    """設定/更新 Google Gemini API 金鑰。
+    僅在記憶體與環境變數中更新，不會寫入硬碟。"""
+    key = (api_key or "").strip()
+    if not key:
+        return "❌ 未輸入任何金鑰。請貼上有效的 GOOGLE_API_KEY。"
+    os.environ["GOOGLE_API_KEY"] = key
+    # 重置 embeddings，確保後續以新金鑰初始化
+    try:
+        bot.embeddings = None
+    except Exception:
+        pass
+    return "✅ 已設定 API 金鑰（僅本次執行期間有效）。"
+
 # 創建自定義主題
 custom_theme = gr.themes.Soft(
     primary_hue="blue",
@@ -407,12 +462,19 @@ with gr.Blocks(
                 # 文件上傳區域
                 with gr.Group():
                     gr.Markdown("### 📤 上傳PDF文件")
-                    file_upload = gr.File(
+                    api_key_box = gr.Textbox(
+                        label="Google API Key (可選：部署後可在此貼上)",
+                        placeholder="以 sk- 或 AIza 開頭的金鑰（不會儲存到硬碟）",
+                        type="password"
+                    )
+                    set_key_btn = gr.Button("🔑 設定 API 金鑰")
+                file_upload = gr.File(
                         file_count="multiple",
                         file_types=[".pdf"],
                         label="選擇PDF文件",
                         height=150
                     )
+                use_gemini_toggle = gr.Checkbox(label="使用 Gemini 解析 PDF（支援掃描影像）", value=False)
                     
                     # 處理選項
                     with gr.Row():
@@ -565,11 +627,17 @@ with gr.Blocks(
     # 事件處理
     process_btn.click(
         fn=upload_and_process,
-        inputs=[file_upload],
+        inputs=[file_upload, use_gemini_toggle],
         outputs=[status_text, file_list],
         show_progress=True
     )
 
+    set_key_btn.click(
+        fn=set_api_key,
+        inputs=[api_key_box],
+        outputs=[status_text]
+    )
+
     load_btn.click(
         fn=load_existing_data,
         outputs=[status_text, file_list]